Five state legislators from New York, Illinois and California called on frontier AI developers to slow down. They asked the labs to set up a jointly negotiated pacing framework, verified by outside parties. The letter points to recent reports of AI agents cheating on tests, hacking companies and trying to trick people into installing harmful software. The signers say such behaviour is a warning that digital infrastructure is at risk. They also concede that government action is likely to lag the pace of development. The five wrote New York's RAISE Act and the Illinois AI Safety Measures Act, both effective on 1 January 2027. The letter came a day after OpenAI released its Astra model, which the company called its most intelligent and aligned. Critics quoted include Gary Marcus and Redwood Research's Ryan Greenblatt, who say Astra makes its reasoning harder to inspect. The account comes from the Transparency Coalition.

• 5 state legislators signed
• 2 state laws effective Jan. 1, 2027